code-time-discipline

Classifies software tasks and gates code edits behind evidence-backed inspection.

Updated Aug 15, 2025
One-click install
npx skills add https://github.com/yehezkieldio/topaz --skill code-time-discipline-yehezkieldio
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: code-time-discipline
Source: https://github.com/yehezkieldio/topaz/tree/main/.agents/skills/code-time-discipline
Command: npx skills add https://github.com/yehezkieldio/topaz --skill code-time-discipline-yehezkieldio

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Engineers and AI agents often make speculative code edits that add complexity, break contracts, or optimize cold paths. This Skill provides a decision gate that classifies each task as code time, measure first, report only, or not code time before any file is changed. ## Core Features & Use Cases - Decision Gate: Classifies tasks into code time, measure first, report only, or not code time based on source inspection and correctness boundaries. - Evidence Standard: Requires every acted-on finding to document location, trigger, cost, growth, evidence level, recommendation, and a P0/P1/P2 priority. - Editing and Verification Rules: Enforces correctness-preserving patches, scoped changes, and a minimal verification gate set of tests, typecheck, lint, and benchmarks. - Use Case: When asked to optimize a slow endpoint, the Skill first inspects call sites, query plans, and tests, then either applies a focused P1 fix with verification or reports findings without editing. ## Quick Start Use the code-time-discipline skill to decide whether this optimization request is code time and only make evidence-backed changes.

Frequently Asked Questions about code-time-discipline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I decide whether to refactor code or leave it alone?

Classify the task using a decision gate: edit only when source inspection shows a real bug, scaling risk, or contract drift with a correctness-preserving patch available. If the finding is speculative or the path is cold, report findings instead of editing.

How to avoid speculative optimizations when improving code performance?

Require an evidence standard for every finding: location, trigger, cost, growth, evidence level, and priority. Act only on P0 actual bottlenecks and justified P1 issues, and measure first with benchmarks, profiles, or query plans when inspection alone is insufficient.

When should I measure performance before editing code?

Measure first when the suspected issue is non-trivial, performance-sensitive, concurrency-sensitive, or contract-sensitive and source inspection alone is not enough. Use existing tests, benchmarks, profiles, logs, or query plans before changing files.

What verification should run after a performance fix?

Run the smallest meaningful gate set: unit or integration tests for the touched package, typecheck and lint for typed code, and benchmarks or query plans for non-trivial improvements. If a gate cannot run, state the exact blocker rather than claiming verification.

When should a code change request be declined?

Decline editing when the edit would weaken validation, authorization, security, transactionality, ordering, public API contracts, or maintainability, or when the request is answered by explanation or review findings. State the classification directly and provide the next useful action.